Skip to content

Type 1 diabetes — insulin therapy and technology

TL;DR — Physiologic insulin replacement requires basal coverage plus meal and correction dosing; CGM has made glucose exposure and variability visible, and automated insulin delivery (AID) now provides the largest consistent technology gain. In the six-month iDCL randomized trial, closed loop increased time 70–180 mg/dL from 61% to 71%, an adjusted 11-point advantage (95% CI 9–14), while reducing time below 70 mg/dL by 0.88 points (Brown 2019, PMID 31618560). CGM also benefits injection users: DIAMOND reduced HbA1c by an adjusted 0.6 percentage points (95% CI 0.3–0.8) at 24 weeks (Beck 2017, PMID 28118453). Devices reduce but do not remove carbohydrate estimation, infusion failure, alarm burden, training, supply, and equity constraints.

Components of contemporary therapy

Layer Function Principal failure mode
Basal insulin Suppress hepatic glucose output between meals/overnight Excess causes fasting hypoglycemia; interruption can cause ketosis
Prandial insulin Match carbohydrate and meal kinetics Timing and composition mismatch
Correction insulin Treat unanticipated hyperglycemia Insulin stacking
CGM Continuous glucose, trend and alerts Compression, lag, adhesion, alarm fatigue
Pump Programmable rapid-acting insulin delivery Set occlusion/dislodgement removes basal insulin
AID algorithm Adjusts insulin from CGM Still limited by insulin kinetics and missed meals
Data platform Pattern review and remote sharing Interoperability, privacy, unequal broadband/device access

Glycemic outcome framework

CGM metric Consensus target for most adults Interpretation
Time 70–180 mg/dL >70% About 16 h 48 min/day in range
Time <70 mg/dL <4% Limits overall biochemical hypoglycemia
Time <54 mg/dL <1% Limits clinically important low exposure
Time >180 mg/dL <25% Hyperglycemia burden
Time >250 mg/dL <5% Marked hyperglycemia/ketosis-risk signal

These targets are consensus treatment goals rather than natural thresholds; pregnancy, frailty, hypoglycemia risk, and individual circumstances require different targets (Battelino 2019, PMID 31177185).

Continuous glucose monitoring

DIAMOND randomized 158 adults on multiple daily injections with baseline HbA1c 7.5–9.9%. At 24 weeks, HbA1c fell 1.0 points with CGM versus 0.4 with usual care; the adjusted difference was −0.6 points (95% CI −0.8 to −0.3), and median time below 70 mg/dL was 43 versus 80 minutes/day (Beck 2017, PMID 28118453).

Flash monitoring also reduced hypoglycemia in well-controlled injection users in the IMPACT subgroup, but its enrollment excluded impaired awareness and recent DKA, limiting inference for those at highest acute risk (Oskarsson 2018, PMID 29273897).

CGM benefit Evidence signal Residual issue
Lower HbA1c Randomized evidence in injection users Benefit depends on sustained wear and response to data
Less hypoglycemia RCT and sensor metrics Severe events are too rare for many trials to power
Trend prediction Alerts before thresholds False/repetitive alarms can impair sleep
Remote follow Useful for children/dependents Surveillance may create conflict or loss of autonomy
Retrospective patterns Enables dose refinement Data overload without clinical support

Pumps and automated insulin delivery

AID joins CGM, pump, and control algorithm. “Hybrid” systems still require meal announcement; “fully closed loop” seeks to remove this input. The iDCL trial enrolled 168 people aged 14–71 and found an 11-point TIR advantage (95% CI 9–14) over sensor-augmented pump therapy at six months (Brown 2019, PMID 31618560).

In 74 children aged 1–7, two 16-week crossover periods produced 8.7 points more TIR (95% CI 7.4–9.9) and HbA1c 0.4 points lower with closed loop, without a significant between-treatment difference in time below 70 mg/dL (Ware 2022, PMID 35045227).

The insulin-only bionic pancreas trial in 161 adults reduced HbA1c by an adjusted 0.5 points (95% CI 0.3–0.6) and increased TIR by 11 points (2.6 h/day) over 13 weeks; severe hypoglycemia occurred in 7/107 intervention and 2/54 control participants (Kruger 2022, PMID 36173236).

Meta-analyses support AID gains across outpatient pediatric trials and real-world cohorts, while emphasizing heterogeneity between algorithms, comparators, baseline control, and follow-up (Zeng 2023, PMID 38011519; Yang 2024, PMID 38888056; Di Molfetta 2024, PMID 39298688).

What AID does not automate

  • Infusion-set inspection and replacement.
  • Ketone testing and injected backup insulin when delivery failure is suspected.
  • Carbohydrate estimation for most commercial systems.
  • Exercise planning and management of delayed hypoglycemia.
  • Sick-day decisions and emergency escalation.
  • Supply ordering, charging, calibration where required, and travel redundancy.
  • Psychological adaptation to alarms, visibility, and device wear.

Tight control from diagnosis using intensive technology did not preserve mixed-meal C-peptide in newly diagnosed children, separating glycemic automation from immune disease modification (McVean 2023, PMID 36826834).

Safety engineering

Hazard Early signal Required system response
Infusion interruption Rising glucose despite corrections; ketones Change set, check ketones, use backup injection plan
Sensor artifact Reading inconsistent with symptoms Confirm glucose before high-consequence action when feasible
Insulin stacking Rapid fall after repeated corrections Account for active insulin
Exercise Falling trend during/after activity Adjust target, insulin and carbohydrate prospectively
Alarm fatigue Alerts ignored or disabled Rationalize thresholds and address recurrent cause
Cyber/data failure Loss of display or integration Maintain independent meter and written backup plan

Access and equity

In the 2016–2018 T1D Exchange, only 17% of youth and 21% of adults met then-current HbA1c goals, and technology use varied by race and socioeconomic status (Foster 2019, PMID 30657336). Young-adult cohorts and safety-net data document racial-ethnic disparities in pump and CGM use and outcomes (Agarwal 2021, PMID 33155826; Fantasia 2021, PMID 33719610).

An algorithm cannot compensate for interrupted insulin supply, unaffordable consumables, absent training, or lack of replacement equipment. Trials often provide devices and intensive support, so implementation studies must report discontinuation, wear, coverage, and missing-data patterns rather than efficacy alone.

Magnitude and limits of technology effects

A network meta-analysis of 28 randomized trials found all hybrid closed-loop systems increased TIR and reduced time below range versus insulin therapy without CGM. Apparent system ranking was uncertain: MiniMed 780G exceeded Control-IQ by 5.1 points (95% CI 0.68–9.52; low certainty) and CamAPS FX by 8.94 (4.35–13.54; low certainty), while severe hypoglycemia and DKA did not differ (Di Molfetta 2024, PMID 39298688).

Evidence setting TIR effect HbA1c effect Boundary
22 trials, 12–96 weeks +10.87 points (95% CI 9.38–12.37) −0.37% (−0.49 to −0.26) Heterogeneity I² 87% and 77%
Youth RCTs, 786–901 participants +11.5 points (9.3–13.7) −0.41% (−0.58 to −0.25) Only two trials reported quality of life
Real-world before–after, 101,704 users +11.61 points (10.47–12.76) −0.42% (−0.47 to −0.37) Selection and regression to mean
Young children RCT, n=102 +12.4 points (9.5–15.3) Secondary improvement One DKA and two severe lows with closed loop

The long-duration RCT synthesis included 2,376 participants and reduced distress but not treatment satisfaction or fear; rare adverse-event inference remained weak (Godoi 2023, PMID 37759290). The youth meta-analysis showed nighttime TIR +19.7 points (17.0–22.4) but insufficient patient-reported outcomes (de Visser 2025, PMID 40920375). Before–after data support effectiveness across age groups but cannot establish comparative safety (Yang 2024, PMID 38888056).

Baseline control changes absolute benefit

DIAMOND enrolled injection-treated adults with baseline HbA1c 7.5–9.9%; CGM produced a −0.6% adjusted HbA1c difference (95% CI −0.8 to −0.3) and reduced median time <70 mg/dL from 80 to 43 minutes/day, while severe events were 2 versus 2 (Beck 2017, PMID 28118453). Across 22 CGM RCTs, mean HbA1c benefit was −0.23% (−0.35 to −0.10) but −0.43% (−0.55 to −0.30) when baseline HbA1c exceeded 8%; severe hypoglycemia and DKA differences were not demonstrated (Teo 2022, PMID 35141761).

ADAPT deliberately enrolled adults with HbA1c ≥8% using injections plus scanned CGM. At six months, HbA1c fell 1.54% with advanced hybrid closed loop versus 0.20% with continued injection/CGM therapy, showing that transition effects can be much larger in a selected high-HbA1c population than in already well-controlled trial cohorts (Choudhary 2022, PMID 36058207).

Insulin kinetics remain the bottleneck

Ultra-rapid lispro was noninferior to lispro for HbA1c in a 432-person pump trial and lowered meal-test glucose by 24.1 mg/dL (95% CI 12.2–36.0) at one hour and 27.8 mg/dL (13.0–42.6) at two hours, but caused more treatment-emergent events (60.5% vs 44.7%), driven by infusion-site reactions (Warren 2021, PMID 33687783). Faster absorption improved a component of the system without eliminating site or meal burdens.

In an 18-adult missed-bolus crossover, fully closed loop with ultra-rapid versus standard lispro produced TIR 49.3% versus 39.9%, but p=0.072; the eight-hour experiment was too small to establish superiority (Thabit 2025, PMID 40815068). Meal-independent automation is therefore not yet equivalent to physiological insulin.

Failure modes and human factors

Failure mode Mechanism Consequence
Interstitial lag Tissue glucose follows falling blood glucose During aerobic exercise, mean lag 12±11 minutes and MARD 13%
Infusion interruption No long-acting depot in pump users Ketosis can develop rapidly
Missed/late meal announcement Subcutaneous absorption trails carbohydrate Postprandial hyperglycemia despite algorithm correction
Alarm fatigue High alert volume and sleep interruption Silencing, discontinuation, or delayed response
Data/connectivity loss Phone, server, transmitter, or upload failure Degraded automation and fragmented review
Supply/coverage interruption Administrative and financial barriers Rationing, device gaps, DKA risk

Exercise data show why symptom–sensor mismatch requires a backup capillary check: CGM lagged the blood-glucose fall by 12±11 minutes in 17 adults (Zaharieva 2019, PMID 31059282). A connected pen cap reduced time >250 mg/dL by 4.8 points (95% CI 0.1–9.5) in a 41-person eight-week trial, but did not change satisfaction; this is promising low-burden support, not an AID substitute (Sebastian-Valles 2026, PMID 41634174).

Access failure is biological risk. Qualitative interviews with 30 adults/caregivers who sought emergency supplies through social media described coverage delays, rationing, and DKA after interrupted access (Sklar 2026, PMID 41996632). Such selected qualitative evidence cannot estimate prevalence but identifies failure mechanisms omitted from device efficacy trials.

Pregnancy is a separate control problem

The CIRCUIT trial randomized 91 pregnant participants. Closed loop raised time in the pregnancy-specific 63–140 mg/dL range from 50.3% to 65.4%, an adjusted +12.5 points (95% CI 9.5–15.6); severe hypoglycemia occurred once and DKA twice versus once with standard care (Donovan 2025, PMID 41134589). Pregnancy-specific targets, changing insulin resistance, and fetal outcomes prevent direct extrapolation from general adult algorithms.

Open questions

  • Can meal-announcement-free systems match hybrid systems safely across high-fat meals, exercise, illness, and puberty? (Di Molfetta 2024, PMID 39298688)
  • Which onboarding and remote-support components cause durable benefit rather than short-term trial adherence? (Yang 2024, PMID 38888056)
  • Do AID systems reduce severe hypoglycemia and DKA at population scale, not only improve surrogate CGM metrics? (Zeng 2023, PMID 38011519)
  • Which reimbursement designs close rather than widen technology disparities? (Agarwal 2021, PMID 33155826; Fantasia 2021, PMID 33719610)
  • How should algorithms disclose adaptation, data use, and failure behavior to users and clinicians?

References

  1. Diabetes Control and Complications Trial Research Group. Effect of intensive treatment on long-term complications. N Engl J Med. 1993;329:977-986. PMID 8366922
  2. Battelino T, et al. Clinical Targets for Continuous Glucose Monitoring Data Interpretation. Diabetes Care. 2019;42:1593-1603. PMID 31177185
  3. Beck RW, et al. Effect of Continuous Glucose Monitoring on Glycemic Control in Adults Using Insulin Injections. JAMA. 2017;317:371-378. PMID 28118453
  4. Oskarsson P, et al. Impact of flash glucose monitoring on hypoglycaemia in adults using injections. Diabetologia. 2018;61:539-550. PMID 29273897
  5. Brown SA, et al. Six-Month Randomized, Multicenter Trial of Closed-Loop Control. N Engl J Med. 2019;381:1707-1717. PMID 31618560
  6. Ware J, et al. Randomized Trial of Closed-Loop Control in Very Young Children. N Engl J Med. 2022;386:209-219. PMID 35045227
  7. Kruger D, et al. Multicenter Randomized Trial of the Insulin-Only Bionic Pancreas in Adults. Diabetes Technol Ther. 2022;24:697-711. PMID 36173236
  8. Zeng B, et al. Automated Insulin Delivery Systems in Children and Adolescents. Diabetes Care. 2023. PMID 38011519
  9. Yang Q, et al. Real-world glycaemic outcomes of automated insulin delivery. Diabetes Obes Metab. 2024. PMID 38888056
  10. Di Molfetta S, et al. Efficacy and Safety of Different Hybrid Closed Loop Systems. Diabetes Metab Res Rev. 2024. PMID 39298688
  11. McVean J, et al. Effect of Tight Glycemic Control on Pancreatic Beta Cell Function. JAMA. 2023. PMID 36826834
  12. Foster NC, et al. State of Type 1 Diabetes Management and Outcomes from T1D Exchange. Diabetes Technol Ther. 2019;21:66-72. PMID 30657336
  13. Agarwal S, et al. Racial-Ethnic Disparities in Diabetes Technology Use Among Young Adults. Diabetes Technol Ther. 2021;23:306-313. PMID 33155826
  14. Fantasia KL, et al. Racial Disparities in Diabetes Technology Use and Outcomes in a Safety-Net Hospital. J Diabetes Sci Technol. 2021;15:1010-1017. PMID 33719610
  15. Godoi A, et al. Glucose control and psychosocial outcomes with automated insulin delivery for 12 to 96 weeks. Diabetol Metab Syndr. 2023;15:190. PMID 37759290
  16. de Visser HS, et al. Automated Insulin Delivery Systems in Children and Adolescents: systematic review and meta-analysis. JAMA Pediatr. 2025;179:1162-1171. PMID 40920375
  17. Teo E, et al. Effectiveness of continuous glucose monitoring in type 1 diabetes: systematic review and meta-analysis. Diabetologia. 2022;65:604-619. PMID 35141761
  18. Choudhary P, et al. Advanced hybrid closed loop versus conventional treatment in adults with type 1 diabetes. Lancet Diabetes Endocrinol. 2022;10:720-731. PMID 36058207
  19. Warren M, et al. Improved postprandial glucose control with ultra rapid lispro versus lispro in pump therapy. Diabetes Obes Metab. 2021;23:1552-1561. PMID 33687783
  20. Thabit H, et al. Fully closed-loop control with ultra-rapid versus standard insulin lispro. Diabet Med. 2025;42:e70122. PMID 40815068
  21. Zaharieva DP, et al. Lag Time Remains with Newer Real-Time CGM During Aerobic Exercise. Diabetes Technol Ther. 2019;21:313-321. PMID 31059282
  22. Sebastian-Valles F, et al. Connected smart pen cap in adults with type 1 diabetes: randomized trial. Diabetologia. 2026;69:1191-1204. PMID 41634174
  23. Sklar J, et al. Barriers to type 1 diabetes medication access and social media support. Diabet Med. 2026;43:e70313. PMID 41996632
  24. Donovan LE, et al. Closed-Loop Insulin Delivery in Type 1 Diabetes in Pregnancy: CIRCUIT. JAMA. 2025;334:2176-2185. PMID 41134589
  25. Wadwa RP, et al. Trial of Hybrid Closed-Loop Control in Young Children with Type 1 Diabetes. N Engl J Med. 2023;388:991-1001. PMID 36920756